Jev 让印度消费级 AI 产品算得过账
印度 ARPU 只有每月 50–100 卢比，养不起每次交互都调前沿模型；作者认为 Jev 的低成本让电商助手、记账助手、学习 App 这类高频决策产品变得可行。
jev can transform indian consumer products. we have a structural problem in consumer AI: ARPU is low, but useful AI products often require hundreds of decisions per user session. at Rs 50–100 a month, you can't afford to call a frontier model every time you need to decide what a user should see, buy or do next. so a lot of products simply don't get built. jev changes the equation. a personal commerce assistant that watches what you browse, learns your preferences and decides what to show you next. a financial assistant that continuously classifies transactions, detects unusual behaviour, and tells you what needs attention. a learning app that makes hundreds of tiny decisions about what a student should practise next, which explanation to give, and when to intervene. these products need intelligence everywhere. the old economics forced you to ration it. in India, lower cost means products that were economically impossible can now exist.